Controlling data consistency guarantees in storage apparatus

• US 8,341,458 B2
• Filed: 07/14/2005
• Issued: 12/25/2012
• Est. Priority Date: 07/17/2004
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Butterworth, Henry E., Fuente, Carlos F.
• Primary Examiner(s): Bonzo, Bryce P
• Assistant Examiner(s): Arcos, Jeison C
• Application Number:
  US11/181,492
• Publication Number:
  US 20060015507A1
• Time in Patent Office: 2,721 Days
• Field of Search: 714/6, 714/7, 714/3.6, 714/707, 709/225, 711/111, 711/203, 707/622, 707/656, 707/662
• US Class Current: 714/6.3
• CPC Class Codes: G06F 11/2064 while ensuring consistency
• CPC Class Codes: G06F 11/2071 using a plurality of controllers
• CPC Class Codes: G06F 2201/82 Solving problems relating to consistency ensuring consistency in mirrored systems G06F11/2064
• CPC Class Codes: G06F 2201/835 Timestamp

Abstract

An apparatus, method and computer program operable in a network controls consistency guarantees of a plurality of data copies in a storage apparatus, and includes: a lease control component for extending a lease to the storage apparatus; and a consistency freeze/thaw component responsive to a failure indication for initiating a consistency freeze action at the storage apparatus prior to expiration of the current group lease period. It may further include a timer component for waiting the apparatus for a predetermined period after the consistency freeze action, where the consistency freeze/thaw component is operable to initiate a thaw action at the storage apparatus.

Claims

1. A peer-to-peer storage system comprising:

a plurality of peer-to-peer storage nodes, each peer-to-peer storage node storing a copy of data such that the copy of the data has a consistency guarantee across the peer-to-peer storage system in that regardless of which peer-to-peer storage node from which the copy of the data is retrieved, the copy of the data is identical;

a consistency freeze/thaw component at each peer-to-peer storage node, responsive to an indication of a communication failure between any two peer-to-peer storage nodes of the plurality of peer-to-peer storage nodes, to initiate a consistency freeze action in which the copy of the data is maintain at a current state and is not permitted to change,

wherein the consistency freeze action is initiated responsive to the indication of the communication failure prior to expiration of a current group lease period concerning at least the copy of the data,

wherein the consistency freeze action is initiated by the consistency freeze/thaw component at each peer-to-peer storage node under a timing constraint related to local clocks at the peer-to-peer storage nodes without a common system clock of the peer-to-peer storage system as a whole,

and wherein the timing constraint takes into account a worst-case time difference among the local clocks due at least to varying local clock rate drifts and local lock rate jitter.

2. The peer-to-peer storage system of claim 1, wherein the indication of the communication failure comprises a failure to receive a given polling message by a first peer-to-peer storage node of the plurality of peer-to-peer storage nodes from a second peer-to-peer storage node of the plurality of peer-to-peer storage nodes.

3. The peer-to-peer storage system of claim 2, wherein the current group lease period is related to a period T(p) at which polling messages are communicated among the peer-to-peer storage nodes.

4. The peer-to-peer storage system of claim 3, wherein the indication of the communication failure is a failure of the first peer-to-peer storage node to receive the given polling message from the second peer-to-peer storage node at a current polling time T(n),

wherein the timing constraint comprises that the consistency freeze action is initiated so that the consistency freeze action is completed before a polling time T(n−

2) plus the current group lease period T(1),

and wherein the polling time T(n−

2) is a polling time immediately prior to a polling time T(n−

1) that is a polling time immediately prior to the current polling time T(n).

5. The peer-to-peer storage system of claim 4, wherein the timing constraint further comprises that the current group lease period T(1) is greater than two times the period T(p) at which the polling messages are communicated among the peer-to-peer storage nodes.

6. A peer-to-peer storage node of a peer-to-peer storage system including a plurality of peer-to-peer storage nodes including the peer-to-peer storage node, comprising:

a storage device to store a copy of data such that the copy of the data has a consistency guarantee across the peer-to-peer storage system in that regardless of which peer-to-peer storage node of the peer-to-peer storage system from which the copy of the data is retrieved, the copy of the data is identical;

a consistency freeze/thaw component, responsive to an indication of a communication failure between the peer-to-peer storage node and another peer-to-peer storage node of the peer-to-peer storage system, to initiate a consistency freeze action in which the copy of the data is maintain at a current state and is not permitted to change,

wherein the consistency freeze action is initiated responsive to the indication of the communication failure prior to expiration of a current group lease period concerning at least the copy of the data,

wherein the consistency freeze action is initiated by the consistency freeze/thaw component at the peer-to-peer storage node under a timing constraint related to local clocks at the peer-to-peer storage nodes without a common system clock of the peer-to-peer storage system as a whole,

and wherein the timing constraint takes into account a worst-case time difference among the local clocks due at least to varying local clock rate drifts and local lock rate jitter.

7. The peer-to-peer storage node of claim 6, wherein the indication of the communication failure comprises a failure to receive a given polling message by the peer-to-peer storage node from the another peer-to-peer storage node.

8. The peer-to-peer storage node of claim 7, wherein the current group lease period is related to a period T(p) at which polling messages are communicated among the peer-to-peer storage nodes.

9. The peer-to-peer storage node of claim 8, wherein the indication of the communication failure is a failure of the peer-to-peer storage node to receive the given polling message from the another peer-to-peer storage node at a current polling time T(n),

wherein the timing constraint comprises that the consistency freeze action is initiated so that the consistency freeze action is completed before a polling time T(n−

2) plus the current group lease period T(1),

and wherein the polling time T(n−

2) is a polling time immediately prior to a polling time T(n−

1) that is a polling time immediately prior to the current polling time T(n).

10. The peer-to-peer storage system of claim 9, wherein the timing constraint further comprises that the current group lease period T(1) is greater than two times the period T(p) at which the polling messages are communicated among the peer-to-peer storage nodes.

11. A method comprising:

at each peer-to-peer storage node of a plurality of peer-to-peer storage nodes of a peer-to-peer storage system, storing a copy of data such that the copy of the data has a consistency guarantee across the peer-to-peer storage system in that regardless of which peer-to-peer storage node from which the copy of the data is retrieved, the copy of the data is identical; and

at each peer-to-peer storage node, responsive to an indication of a communication failure between any two peer-to-peer storage nodes of the plurality of peer-to-peer storage nodes, initiating a consistency freeze action in which the copy of the data is maintain at a current state and is not permitted to change,

wherein the consistency freeze action is initiated responsive to the indication of the communication failure prior to expiration of a current group lease period concerning at least the copy of the data,

wherein the consistency freeze action is initiated by the consistency freeze/thaw component at each peer-to-peer storage node under a timing constraint related to local clocks at the peer-to-peer storage nodes without a common system clock of the peer-to-peer storage system as a whole,

and wherein the timing constraint takes into account a worst-case time difference among the local clocks due at least to varying local clock rate drifts and local lock rate jitter.

12. The method of claim 11, wherein the indication of the communication failure comprises a failure to receive a given polling message by a first peer-to-peer storage node of the plurality of peer-to-peer storage nodes from a second peer-to-peer storage node of the plurality of peer-to-peer storage nodes.

13. The method of claim 12, wherein the current group lease period is related to a period T(p) at which polling messages are communicated among the peer-to-peer storage nodes.

14. The method of claim 13, wherein the indication of the communication failure is a failure of the first peer-to-peer storage node to receive the given polling message from the second peer-to-peer storage node at a current polling time T(n),

wherein the timing constraint comprises that the consistency freeze action is initiated so that the consistency freeze action is completed before a polling time T(n−

2) plus the current group lease period T(1),

and wherein the polling time T(n−

2) is a polling time immediately prior to a polling time T(n−

1) that is a polling time immediately prior to the current polling time T(n).

15. The method of claim 14, wherein the timing constraint further comprises that the current group lease period T(1) is greater than two times the period T(p) at which the polling messages are communicated among the peer-to-peer storage nodes.

16. An article of manufacture for a peer-to-peer storage node of a peer-to-peer storage system including a plurality of peer-to-peer storage nodes including the peer-to-peer storage node, comprising:

a non-transitory computer-readable medium; and

means in the medium for, responsive to an indication of a communication failure between the peer-to-peer storage node and another peer-to-peer storage node of the peer-to-peer storage system, initiating a consistency freeze action in which the copy of the data is maintain at a current state and is not permitted to change,

wherein the consistency freeze action is initiated responsive to the indication of the communication failure prior to expiration of a current group lease period concerning at least the copy of the data,

wherein the consistency freeze action is initiated by the consistency freeze/thaw component at the peer-to-peer storage node under a timing constraint related to local clocks at the peer-to-peer storage nodes without a common system clock of the peer-to-peer storage system as a whole,

wherein the timing constraint takes into account a worst-case time difference among the local clocks due at least to varying local clock rate drifts and local lock rate jitter,

and wherein the peer-to-peer storage node is to store a copy of data such that the copy of the data has a consistency guarantee across the peer-to-peer storage system in that regardless of which peer-to-peer storage node of the peer-to-peer storage system from which the copy of the data is retrieved, the copy of the data is identical.

17. The article of manufacture of claim 16, wherein the indication of the communication failure comprises a failure to receive a given polling message by the peer-to-peer storage node from the another peer-to-peer storage node.

18. The article of manufacture of claim 17, wherein the current group lease period is related to a period T(p) at which polling messages are communicated among the peer-to-peer storage nodes.

19. The article of manufacture of claim 18, wherein the indication of the communication failure is a failure of the peer-to-peer storage node to receive the given polling message from the another peer-to-peer storage node at a current polling time T(n),

wherein the timing constraint comprises that the consistency freeze action is initiated so that the consistency freeze action is completed before a polling time T(n−

2) plus the current group lease period T(1),

and wherein the polling time T(n−

2) is a polling time immediately prior to a polling time T(n−

1) that is a polling time immediately prior to the current polling time T(n).

20. The article of manufacture of claim 19, wherein the timing constraint further comprises that the current group lease period T(1) is greater than two times the period T(p) at which the polling messages are communicated among the peer-to-peer storage nodes.